EU says people should be held responsible for ‘bad’ AI

21 Oct 2020

Image: © scharfsinn86/Stock.adobe.com

The European Parliament has voted in favour of a number of proposals for regulating AI with regards to ethics, liability and intellectual property.

Three proposals aiming to lay the groundwork for future AI regulations in the EU were approved by the European Parliament during a vote yesterday (20 October).

The first initiative, proposed by MEP Iban García del Blanco, urged the EU to create a legal framework outlining the ethical principles and legal obligations to be followed when deploying and using AI, robotics and related technologies. This would include software, algorithms and data.

MEPs voted 559 in favour and 44 against, with 88 abstentions, for a set of recommendations that would see future laws made in accordance with several principles. These include human-centric and human-made AI; transparency and accountability; safeguards against bias and discrimination; and the right to redress.

Under García del Blanco’s proposal, high-risk AI technologies – such as those with self-learning capabilities – would need to allow for human oversight at any time. If it’s deemed that an AI could be dangerous and breaches ethical guidelines, its ability to self-learn should be disabled to allow the human to take over.

Infographic showing the impact of AI on patient applications and its impact on jobs.

Image: European Commission

AI insurance

A second proposal put forward by MEP Axel Voss would make those operating high-risk AI strictly liable for any resulting damage. With 626 votes in favour of the initiative and just 25 against, the proposal would apply to physical or virtual AI activity that harms or damages life, health, physical integrity or property. This would also apply if the AI contributed to “verifiable economic loss”.

This would result in a situation where high-risk AI designers would be required to take out insurance, similar to how a car owner requires insurance before taking to the road.

A third proposal put forward by MEP Stéphane Séjourné called for an effective intellectual property rights system and safeguards for the EU’s patent system. This, the proposal said, would help create legal precedent for distinguishing between AI-assisted human creations and AI-generated creations. It was adopted with 612 votes in favour, 66 against and 12 abstentions.

MEPs argued that AI should not have legal personality and that intellectual property rights should only be awarded to humans.

Need for trustworthy AI

Ahead of a European Commission legislative proposal on AI, expected in early 2021, a special committee was established to gauge the impact of AI on the EU economy.

The new committee chair, Dragoș Tudorache, said: “Europe needs to develop AI that is trustworthy, eliminates biases and discrimination and serves the common good, while ensuring business and industry thrive and generate economic prosperity.”

European Commission president Ursula von der Leyen ordered her top deputies to come up with a coordinated European approach to AI and data strategy in the first 100 days after she took office in December 2019. In an op-ed, von der Leyen said that despite the EU drawing up a precautionary framework regarding AI, she remains a “tech optimist” but that “digital transformation cannot be left to chance”.

Colm Gorey was a senior journalist with Silicon Republic

editorial@siliconrepublic.com